Aakash Gupta: Seven AI competencies for product managers — Berkeley ProductCon talk

What the video covers

Aakash Gupta — a writer and advisor focused on AI for product managers — delivered this keynote at Berkeley’s ProductCon. The talk structures PM AI skills into seven areas: prompting, AI copilots, AI agents, prototyping, AI-assisted discovery, building AI features, and AI analysis. The recording runs roughly an hour and covers each area in sequence with tool recommendations, providing a framework that teams can use to identify which competencies are developed and which need work.

Who it’s for

Product managers who have experimented with AI tools but have not organized their approach into a system they can repeat or explain. Also useful for team leads and heads of product looking to establish a shared vocabulary and baseline skill expectations. The talk mentions that companies like Zapier and Shopify have begun evaluating PM performance on AI proficiency, so there is a practical career dimension alongside the operational one.

Key takeaways

1. Prompting skill differs more across practitioners than most teams realize. Gupta uses the analogy of Excel proficiency — the gap between someone who enters data manually and someone who builds complex models is enormous, yet both could be described as “using Excel.” The RTF framework (Role, Task, Format) is presented as a useful baseline. Only 5% of PMs maintain organized prompt libraries, which compounds the skill gap over time as those who do iterate their prompts while others restart from scratch.

2. System prompts can eliminate sycophancy from AI feedback. A technique Gupta calls “Absolute Mode” — instructing the model upfront to skip positive reinforcement and give direct, critical responses — produces more useful results on PRDs, feature briefs, and strategy documents. Without this kind of instruction, AI models tend to validate whatever you write rather than identify what is weak or missing.

3. Dictation tools roughly double the speed of getting material into a model. Typing is a bottleneck when working through discovery notes, call transcripts, or rough strategic thinking. Tools like Wispr Flow and Speechify convert speech to text fast enough that they change the practical economics of using AI for analysis and drafting during an active workday.

4. A seven-competency framework makes AI adoption measurable rather than vague. Rather than asking whether a team uses AI, Gupta’s structure makes it possible to identify which specific areas — agents, prototyping, analysis — are underdeveloped. This is useful both for personal skill assessment and for setting clearer expectations in hiring and performance conversations.

Worth watching if…

You use AI tools in your PM work but find the results inconsistent and hard to explain to colleagues, or if you are building out AI skill expectations for a product organization and want a framework with enough specifics to act on.